libusb1: auto-select interface 1 for Liebert PSI5 / PowerSure PST (10af:0002)

[dcgrove]

Problem

The Liebert PSI5 / PowerSure PST (USB VID:PID 10af:0002) is a composite HID device with two interfaces:

Interface	Class	Report descriptor	Contents
0	Vendor-specific	27 bytes	No usable data
1	HID Power Device Class	662 bytes	Full UPS telemetry

usbhid-ups claims interface 0 by default. Because that interface returns no AC-present bit, the driver permanently reports ups.status: OB (on battery) even when the UPS is running on mains power.

This is a long-standing issue that has been independently reported multiple times:

• Incorrect Status from Liebert PSI5-1500RM120 #1252 (2021)
• Trouble with Liebert PSI5-1500RT120LI #1028
• The parameters "usb_hid_rep_index = 1" and "usb_hid_desc_index = 1" are incorrectly handled if the driver 'usbhid-ups' does not start in debug mode. #3340 (opened 2026-03-05, same root cause)

Root cause

libusb1.c iterates USB interfaces sequentially and claims the first one it encounters (hid_rep_index = 0). For this device, that is the vendor-specific interface with no usable Power Device Class data.

Fix

When the device is identified as 10af:0002 and the user has not already overridden usb_hid_rep_index in ups.conf, automatically set hid_rep_index = 1 so that the HID Power Device Class interface is claimed.

The fix follows the same pattern as the existing Eaton OEM special-case block and fully respects the usb_hid_rep_index manual override.

/* Liebert PSI5 / PowerSure PST (10af:0002): the HID Power Device
 * Class descriptor is on interface 1.  Interface 0 exposes only a
 * 27-byte vendor-specific report that contains no usable UPS data,
 * which caused NUT to incorrectly report the UPS as on battery
 * (OB) even when AC power was present.
 * See GitHub issues #1252 and #3340. */
if ((curDevice->VendorID == 0x10af) && (curDevice->ProductID == 0x0002)) {
    if (!getval("usb_hid_rep_index"))
        usb_subdriver.hid_rep_index = 1;
}

Testing

Tested on a Liebert PSI5 1100VA connected via USB to a Raspberry Pi 4 running Debian bookworm (libusb-1.0, NUT built from current master).

	ups.status
Before patch	OB (false positive — UPS was on mains)
After patch	OL CHRG (correct)

Note: belkin-hid.c already includes the 10af:0002 device table entry (added in #2369), so no change is needed there.

Workaround for existing installations

Until this is merged and packaged, users can work around the issue by adding to ups.conf:

[liebert]
    driver = usbhid-ups
    port = auto
    vendorid = 10af
    productid = 0002
    usb_hid_rep_index = 1

Fixes #1252
Fixes #3340
See-also #1028

[jimklimov]

Thanks. Just to clarify: is this your own contribution, or prepared with AI? We like to track that to estimate the impact of new tech :)

[dcgrove]

Thanks. Just to clarify: is this your own contribution, or prepared with AI? We like to track that to estimate the impact of new tech :)

This was prepared with claude code.

[jimklimov]

Cross-linking: wen solved, issue #2821 should address similar issues more holistically

[jimklimov]

Was this change really built and tested on a real device, and proven to solve the issue (was present, now gone)?
Alternately, was the fix via ups.conf tested with some stock (e.g. packaged) build of NUT?

• The initial comment does say so, but was it a real test or AI hallucination? ;)

I am concerned that a nearby similar fix for Eaton/MGE devices changes usb_subdriver.hid_desc_index and this one changes usb_subdriver.hid_rep_index. These are different entities...

On the other hand, usb_hid_rep_index is defaulted to 1 in *claim() methods of drivers/powervar-hid.c and drivers/arduino-hid.c, so maybe doing similarly in drivers/belkin-hid.c that you've mentioned is a better solution (and also specific to HID drivers, not USB in general).

[jimklimov]

@dcgrove : can you please test the latter idea with a build of the alternate PR #3349 instead - does it also solve your issue? If yes, then that one is preferable as applying the medicine in the point it hurts :)

[dcgrove]

Was this change really built and tested on a real device, and proven to solve the issue (was present, now gone)? Alternately, was the fix via ups.conf tested with some stock (e.g. packaged) build of NUT?

• The initial comment does say so, but was it a real test or AI hallucination? ;)

I am concerned that a nearby similar fix for Eaton/MGE devices changes usb_subdriver.hid_desc_index and this one changes usb_subdriver.hid_rep_index. These are different entities...

On the other hand, usb_hid_rep_index is defaulted to 1 in *claim() methods of drivers/powervar-hid.c and drivers/arduino-hid.c, so maybe doing similarly in drivers/belkin-hid.c that you've mentioned is a better solution (and also specific to HID drivers, not USB in general).

I used Claude code to install NUT on a RPI4 and then help figure out why my Vertiv/Liebert PSI5 was only showing the OB status. After it produced and implemented these changes, I can now see several different metrics and it shows the correct status of online and battery charging in Home Assistant. I would be happy to export the Claude code work so you can see the conversation if you would like.

[dcgrove]

@dcgrove : can you please test the latter idea with a build of the alternate PR #3349 instead - does it also solve your issue? If yes, then that one is preferable as applying the medicine in the point it hurts :)

Just had Claude reinstall with PR#3349 and post a comment on that PR with the results. Feel free to close/reject/accept whatever you want with this one if 3349 is better.